clusterperf August 12, 2013
- Collin Lee
Owned by Collin Lee
As of Rev 16142c1
basic.read100 5.1 us read single 100B object with 30B key
basic.readBw100 18.7 MB/s bandwidth reading 100B object with 30B key
basic.read1K 6.9 us read single 1KB object with 30B key
basic.readBw1K 137.6 MB/s bandwidth reading 1KB object with 30B key
basic.read10K 10.4 us read single 10KB object with 30B key
basic.readBw10K 914.1 MB/s bandwidth reading 10KB object with 30B key
basic.read100K 47.2 us read single 100KB object with 30B key
basic.readBw100K 2.0 GB/s bandwidth reading 100KB object with 30B key
basic.read1M 420.8 us read single 1MB object with 30B key
basic.readBw1M 2.2 GB/s bandwidth reading 1MB object with 30B key
basic.write100 15.7 us write single 100B object with 30B key
basic.writeBw100 6.1 MB/s bandwidth writing 100B object with 30B key
basic.write1K 19.9 us write single 1KB object with 30B key
basic.writeBw1K 48.0 MB/s bandwidth writing 1KB object with 30B key
basic.write10K 38.5 us write single 10KB object with 30B key
basic.writeBw10K 247.7 MB/s bandwidth writing 10KB object with 30B key
basic.write100K 235.3 us write single 100KB object with 30B key
basic.writeBw100K 405.3 MB/s bandwidth writing 100KB object with 30B key
basic.write1M 2.2 ms write single 1MB object with 30B key
basic.writeBw1M 436.0 MB/s bandwidth writing 1MB object with 30B key
broadcast 193.3 us broadcast message to 9 slaves
netBandwidth 542.2 MB/s many clients reading from different servers
netBandwidth.max 16.6 MB/s fastest client
netBandwidth.min 12.4 MB/s slowest client
readNotFound 14.5 us read object that doesn't exist
# Gauges impact of asynchronous writes on synchronous writes.
# Write two values. The size of the first varies over trials
# (its size is given as 'firstObjectSize'). The first write is
# either synchronous (if firstWriteIsSync is 1) or asynchronous
# (if firstWriteIsSync is 0). The response time of the first
# write is given by 'firstWriteLatency'. The second write is
# a 100 B object which is always written synchronously (its
# response time is given by 'syncWriteLatency'
# Both writes use a 30 B key.
# Generated by 'clusterperf.py writeAsyncSync'
#
# firstWriteIsSync firstObjectSize firstWriteLatency(us) syncWriteLatency(us)
#----------------------------------------------------------------------------
0 100 16.2 16.8
0 1000 20.6 16.8
0 10000 40.1 17.4
0 100000 238.3 19.6
0 1000000 2228.1 29.7
1 100 17.2 19.3
1 1000 23.2 20.2
1 10000 47.7 22.4
1 100000 251.1 28.0
1 1000000 2215.7 29.6
# RAMCloud multiWrite performance for 100 B objects with 30 byte keys
# located on a single master.
# Generated by 'clusterperf.py multiWrite_oneMaster'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
1 1 1 16.6 16.6
2 1 2 20.2 10.1
3 1 3 22.4 7.5
4 1 4 25.3 6.3
5 1 5 26.9 5.4
6 1 6 27.7 4.6
7 1 7 30.0 4.3
8 1 8 32.0 4.0
9 1 9 33.6 3.7
10 1 10 37.0 3.7
20 1 20 55.8 2.8
30 1 30 73.2 2.4
40 1 40 89.0 2.2
50 1 50 107.4 2.1
60 1 60 123.2 2.1
70 1 70 143.9 2.1
80 1 80 168.2 2.1
90 1 90 181.2 2.0
100 1 100 196.2 2.0
200 1 200 353.6 1.8
300 1 300 516.4 1.7
400 1 400 677.8 1.7
500 1 500 826.9 1.7
600 1 600 987.4 1.6
700 1 700 1142.9 1.6
800 1 800 1300.0 1.6
900 1 900 1465.9 1.6
1000 1 1000 1634.0 1.6
2000 1 2000 3176.1 1.6
3000 1 3000 4812.4 1.6
4000 1 4000 6435.1 1.6
5000 1 5000 7998.5 1.6
# RAMCloud multiRead performance for 100 B objects with 30 byte keys
# located on a single master.
# Generated by 'clusterperf.py multiRead_oneMaster'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
1 1 1 5.6 5.6
2 1 2 7.2 3.6
3 1 3 8.2 2.7
4 1 4 9.5 2.4
5 1 5 10.5 2.1
6 1 6 11.9 2.0
7 1 7 12.7 1.8
8 1 8 13.9 1.7
9 1 9 15.6 1.7
10 1 10 16.1 1.6
20 1 20 23.9 1.2
30 1 30 32.9 1.1
40 1 40 39.9 1.0
50 1 50 50.6 1.0
60 1 60 59.3 1.0
70 1 70 68.9 1.0
80 1 80 74.9 0.9
90 1 90 73.5 0.8
100 1 100 78.5 0.8
200 1 200 127.4 0.6
300 1 300 179.8 0.6
400 1 400 227.1 0.6
500 1 500 279.8 0.6
600 1 600 330.1 0.6
700 1 700 373.5 0.5
800 1 800 444.1 0.6
900 1 900 487.4 0.5
1000 1 1000 550.4 0.6
2000 1 2000 1089.7 0.5
3000 1 3000 1575.1 0.5
4000 1 4000 2052.6 0.5
5000 1 5000 2590.6 0.5
# RAMCloud multiRead performance for 100 B objects with 30 byte keys
# with one object located on each master.
# Generated by 'clusterperf.py multiRead_oneObjectPerMaster'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
1 1 1 5.5 5.5
2 2 1 6.7 3.3
3 3 1 8.0 2.7
4 4 1 9.4 2.4
5 5 1 10.7 2.1
6 6 1 12.3 2.1
7 7 1 13.7 2.0
8 8 1 15.2 1.9
9 9 1 16.8 1.9
10 10 1 17.0 1.7
11 11 1 21.9 2.0
12 12 1 24.9 2.1
13 13 1 28.8 2.2
14 14 1 30.7 2.2
15 15 1 33.8 2.3
16 16 1 32.1 2.0
17 17 1 44.8 2.6
18 18 1 39.5 2.2
# RAMCloud multiRead performance for an approximately fixed number
# of 100 B objects with 30 byte keys
# distributed evenly across varying number of masters.
# Generated by 'clusterperf.py multiRead_general'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
5000 1 5000 2563.5 0.5
5000 2 2500 3673.8 0.7
4998 3 1666 2662.6 0.5
5000 4 1250 2906.8 0.6
5000 5 1000 2682.1 0.5
4998 6 833 3143.6 0.6
4998 7 714 4756.2 1.0
5000 8 625 3765.5 0.8
4995 9 555 3903.6 0.8
5000 10 500 4030.9 0.8
4994 11 454 4420.7 0.9
4992 12 416 3914.3 0.8
4992 13 384 3934.4 0.8
4998 14 357 6303.5 1.3
4995 15 333 3976.7 0.8
4992 16 312 4019.9 0.8
4998 17 294 4036.9 0.8
4986 18 277 4070.8 0.8
# RAMCloud multiRead performance for an approximately fixed number
# of 100 B objects with 30 byte keys
# distributed evenly across varying number of masters.
# Requests are issued in a random order.
# Generated by 'clusterperf.py multiRead_generalRandom'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
5000 1 5000 2891.6 0.6
5000 2 2500 2851.5 0.6
4998 3 1666 2858.4 0.6
5000 4 1250 3033.0 0.6
5000 5 1000 3167.7 0.6
4998 6 833 3337.1 0.7
4998 7 714 4337.0 0.9
5000 8 625 3812.3 0.8
4995 9 555 3798.8 0.8
5000 10 500 5692.7 1.1
4994 11 454 3998.6 0.8
4992 12 416 4140.2 0.8
4992 13 384 4499.4 0.9
4998 14 357 4797.2 1.0
4995 15 333 4707.5 0.9
4992 16 312 4799.1 1.0
4998 17 294 5351.9 1.1
4986 18 277 5860.4 1.2
# Cumulative distribution of time for a single client to read a
# single 100-byte object from a single server. Each line indicates
# that a given fraction of all reads took at most a given time
# to complete.
# Generated by 'clusterperf.py readDist'
#
# Time (usec) Cum. Fraction
#---------------------------
0.00 0.000
4.70 0.000
4.78 0.010
4.79 0.020
4.79 0.030
4.80 0.040
4.80 0.050
4.81 0.060
4.81 0.070
4.82 0.080
4.82 0.090
4.83 0.100
4.83 0.110
4.83 0.120
4.84 0.130
4.84 0.140
4.84 0.150
4.84 0.160
4.85 0.170
4.85 0.180
4.85 0.190
4.85 0.200
4.85 0.210
4.86 0.220
4.86 0.230
4.86 0.240
4.86 0.250
4.86 0.260
4.86 0.270
4.87 0.280
4.87 0.290
4.87 0.300
4.87 0.310
4.87 0.320
4.87 0.330
4.88 0.340
4.88 0.350
4.88 0.360
4.88 0.370
4.88 0.380
4.88 0.390
4.88 0.400
4.88 0.410
4.88 0.420
4.89 0.430
4.89 0.440
4.89 0.450
4.89 0.460
4.89 0.470
4.89 0.480
4.89 0.490
4.89 0.500
4.89 0.510
4.90 0.520
4.90 0.530
4.90 0.540
4.90 0.550
4.90 0.560
4.90 0.570
4.90 0.580
4.90 0.590
4.90 0.600
4.90 0.610
4.91 0.620
4.91 0.630
4.91 0.640
4.91 0.650
4.91 0.660
4.91 0.670
4.92 0.680
4.92 0.690
4.92 0.700
4.93 0.710
4.93 0.720
4.94 0.730
4.95 0.740
4.96 0.750
4.97 0.760
4.98 0.770
4.99 0.780
5.00 0.790
5.01 0.800
5.03 0.810
5.05 0.820
5.07 0.830
5.12 0.840
5.47 0.850
5.52 0.860
5.54 0.870
5.56 0.880
5.57 0.890
5.59 0.900
5.65 0.910
5.91 0.920
6.03 0.930
6.07 0.940
6.10 0.950
6.12 0.960
6.15 0.970
6.29 0.980
7.26 0.990
72.64 0.999
102.97 0.9999
582.38 1.000
# RAMCloud read performance for 100 B objects
# with keys of various lengths.
# Generated by 'clusterperf.py readVaryingKeyLength'
#
# Key Length Latency (us) Bandwidth (MB/s)
#----------------------------------------------------------------------------
1 5.1 0.2
5 5.0 0.9
10 5.1 1.9
15 5.0 2.8
20 4.9 3.9
25 5.0 4.8
30 5.0 5.7
35 5.1 6.6
40 5.2 7.3
45 5.2 8.3
50 5.1 9.4
55 5.1 10.2
60 5.3 10.8
65 5.3 11.7
70 5.5 12.1
75 5.4 13.3
80 5.4 14.2
85 5.4 15.1
90 5.1 16.8
95 5.6 16.3
100 5.5 17.3
200 6.0 31.8
300 6.6 43.5
400 7.0 54.5
500 6.8 70.0
600 6.9 82.4
700 6.9 96.3
800 7.1 108.0
900 7.2 119.1
1000 7.4 129.7
2000 8.3 230.3
3000 8.9 321.3
4000 9.7 394.2
5000 10.2 469.1
6000 10.8 528.3
7000 11.5 580.1
8000 12.2 623.7
9000 13.6 633.1
10000 13.8 693.2
20000 21.9 871.6
30000 29.4 971.8
40000 39.6 962.2
50000 47.9 995.9
60000 56.2 1018.0
# RAMCloud write performance for 100 B objects
# with keys of various lengths.
# Generated by 'clusterperf.py writeVaryingKeyLength'
#
# Key Length Latency (us) Bandwidth (MB/s)
#----------------------------------------------------------------------------
1 15.8 0.1
5 15.7 0.3
10 15.8 0.6
15 15.8 0.9
20 15.7 1.2
25 16.1 1.5
30 15.7 1.8
35 15.6 2.1
40 15.4 2.5
45 15.5 2.8
50 15.8 3.0
55 16.5 3.2
60 16.3 3.5
65 16.0 3.9
70 15.9 4.2
75 16.0 4.5
80 16.1 4.7
85 16.6 4.9
90 17.1 5.0
95 17.3 5.2
100 18.2 5.2
200 18.5 10.3
300 19.4 14.8
400 19.4 19.7
500 20.5 23.2
600 20.3 28.2
700 20.7 32.2
800 20.8 36.7
900 21.5 40.0
1000 22.3 42.7
2000 27.5 69.3
3000 32.5 88.1
4000 36.4 104.8
5000 42.6 111.9
6000 45.9 124.6
7000 50.1 133.3
8000 53.1 143.6
9000 57.2 150.1
10000 59.2 161.0
20000 104.7 182.2
30000 145.4 196.8
40000 201.2 189.6
50000 249.1 191.4
60000 293.8 194.8
# RAMCloud read performance as a function of load (1 or more
# clients all reading a single 100-byte object with 30-byte key
# repeatedly).
# Generated by 'clusterperf.py readLoaded'
#
# numClients readLatency(us) throughput(total kreads/sec)
#----------------------------------------------------------
1 5.3 189
2 5.4 370
3 6.3 473
4 6.3 632
5 7.6 662
6 9.9 604
7 10.6 661
8 12.2 656
9 15.0 598
10 16.9 592
11 17.0 647
12 18.6 645
13 21.9 595
14 21.4 653
15 25.6 587
16 25.5 629
17 26.6 640
18 28.0 643
19 29.2 650
20 30.7 652
# RAMCloud read performance when 1 or more clients read
# 100-byte objects with 30-byte keys chosen at random from
# 10 servers.
# Generated by 'clusterperf.py readRandom'
#
# numClients throughput(total kreads/sec) slowest(ms) reads > 10us
#--------------------------------------------------------------------
1 177 0.59 0.7%
2 324 0.57 0.8%
3 524 0.33 0.3%
4 671 0.64 0.6%
5 741 6.30 1.1%
6 943 0.57 1.4%
7 1103 0.64 1.3%
8 1183 7.09 1.7%
9 1356 0.61 1.6%
10 1514 0.62 1.8%
11 1564 10.47 1.1%
12 1661 10.48 1.9%
13 1763 10.40 1.7%
14 1984 10.40 1.4%
15 2123 10.43 1.5%
16 2130 10.47 2.0%
17 2281 10.47 2.4%
18 2337 10.46 2.6%
19 2726 0.50 3.1%
20 2857 2.51 2.7%
21 2749 13.12 4.0%
22 2452 0.50 4.2%
23 3225 0.54 4.8%
24 3182 3.38 5.3%
25 3366 0.50 6.5%
26 2788 21.82 6.5%
27 3688 0.60 6.8%
28 3730 0.61 7.7%
29 2316 36.45 10.8%
30 3794 0.61 12.0%
31 3839 0.64 13.3%
32 3681 3.73 16.0%
33 3884 2.12 15.3%
34 1980 57.09 13.8%
35 3967 3.30 20.6%
36 4028 3.84 20.3%
37 2472 28.43 22.5%
38 4080 2.74 23.8%
39 4281 2.29 23.3%
40 4367 2.31 24.3%
41 4296 4.26 26.5%
42 3989 11.12 27.9%
43 4225 12.78 26.4%
44 4452 2.56 29.3%
45 4432 2.60 32.4%
46 4471 2.70 32.7%
47 4530 2.44 36.0%
48 4537 2.46 36.2%
49 4231 4.72 38.9%
50 2995 30.69 35.7%
**** server8.rc11.log:
1376433121.896881152 src/FailureDetector.cc:159 in FailureDetector::pingRandomServer default WARNING[19126:3]: Ping timeout to server id 5.0 (locator "infrc:host=192.168.1.105,port=12247")